Wednesday, October 8, 2008

Fifth projection (or yesterday's numbers, today)

Due to technical difficulties, I didn't get yesterday's update posted. Here it is this morning, with another to follow this evening with today's polls. For this projection, I used the following polls:

Angus Reid - Oct 2-3 (0.64 weight)
EKOS - Oct 4-6 (1.00 weight)
Ipsos-Reid - Sep 30-Oct 2 (0.43 weight)
Nanos Research - Oct 4-6 [ON, QC, Atl only] (1.00 weight)
Strategic Counsel - Sep 28-29 [ON, QC only] (0.07 weight)
Harris-Decima -Oct 3-6 [ON, QC, Atl, BC only] (0.93 weight)

National results (change vs. 2006, change vs. last projection)
CPC 136 (+12, -3), Lib 81 (-22, +6), BQ 49 (-2, n/c), NDP 38 (+9, -3), Grn 1 (+1, n/c), Ind 3 (+2, n/c)

Atlantic (change vs. 2006, change vs. last projection)
Lib 18 (-2, +1), CPC 8 (-1, n/c), NDP 4 (+1, -1), Grn 1 (+1, n/c), Ind 1 (+1, n/c)

Quebec (change vs. 2006, change vs. last projection)
BQ 49 (-2, n/c), Lib 13 (+2, n/c), CPC 11 (+1, n/c), Ind 2 (+1, n/c), NDP 0 (n/c, n/c)

Ontario (change vs. 2006, change vs. last projection)
CPC 44 (+4, -4), Lib 43 (-11, +4), NDP 19 (+7, n/c)

Saskatchewan and Manitoba (change vs. 2006, change vs. last projection)
CPC 22 (+2, -1), Lib 3 (-2, +1), NDP 3 (n/c, n/c)

Alberta (change vs. 2006, change vs. last projection)
CPC 28 (n/c, n/c)

British Columbia (change vs. 2006, change vs. last projection)
CPC 23 (+6, +2), NDP 11 (+1, -2), Lib 2 (-7, n/c)

North (change vs. 2006, change vs. last projection)
Lib 2 (n/c, n/c), NDP 1 (n/c, n/c)

A few remarks...

The Conservatives have slipped again, losing 4 seats to Liberals in Ontario bringing that province essentially to a draw. Their losses were however somewhat offset by gains in BC.

The race in Outremont, Quebec remains close, though, for the first time, it hasn't changed hands in a projection. This edition shows the Liberals increasing their lead from 0.3% to 0.37%.

All three new polls in this projection show the Conservatives either tied (1) or behind (2) in Ontario, they retain a 1 seat lead in this projection but that is largely due to their strength in older polling. If the trend continues, the Conservatives are almost certain to lose their Ontario lead in the projection tonight. However, they formed a government with a 14 seat deficit in Ontario last time, unless the bottom really falls out, it seems likely they'll still have some (marginal) gains in Ontario, though the trend would show them still finishing second, just a stronger second than last time.

I'll continue to update daily until October 14. I welcome your suggestions if you think any aspect of the formula needs to be adjusted.

A question for readers (repeated)

There is a slight glitch right now for the projection model in those four ridings where a party is not contesting and endorsing another candidate. The base number for those ridings had the non-contesting party at 0, as they should be, but when the polling data is entered, that number moves, sometimes into negative territory.

*new* No one has answered this, so I wanted to show how it would affect the results, the three options are as follows, with the results as of this projection in brackets:
  1. Disregard the number for the non-contesting party (Central Nova: Grn 36%, NDP 36%, CPC 34%, Lib -5%) [what I have been doing];
  2. Distribute the gain or loss in the same proportion as I did with the 2006 votes (NDP 34%, CPC 33%, Grn 33%); or
  3. Distribute the gain or loss to the candidate the non-contesting party has endorsed? (Central Nova: NDP 35%, CPC 34%, Grn 31%)
NOTE: These projections do not reflect the wishes, nor predictions, of the author. They are based on simple math using the criteria outlined here and here.

2 comments:

Rob said...

could you weight #3? I shouldn't think a Liberal loss would translate completely to a Green loss.

nbpolitico said...

That's what #2 is.

To create the baseline of Green support in the riding, I gave them 60% of the Liberal vote from 2006 and spread the rest around proportionally. In #2, I do the same thing with the gain/loss that the Liberals are getting.